# Theme
my_theme <- theme(plot.title = element_text(size = rel(1.5), colour = "#929292", family = "Cambria", hjust = .5, vjust = .8),
plot.subtitle = element_text(size = rel(1.2), colour = "#929292", family = "Cambria", hjust =.5),
plot.caption = element_text(colour = "#929292", family = "Calibri"),
legend.text = element_text(colour = "#929292", family = "Calibri"),
legend.title = element_text(colour = "#929292", face = "bold", family = "Cambria"),
axis.title.y = element_text(colour = "#929292", size = rel(1.2), family = "Cambria", angle = 0, vjust = .5),
axis.title.x = element_text(colour = "#929292", size = rel(1.2), family = "Cambria"),
axis.text = element_text(colour = "#929292", family = "Calibri"),
legend.box.background = element_rect(linetype = "solid", colour = "#929292"),
panel.background = element_rect(linetype = "solid", size = .5, colour = "#929292", fill = "#F1F1F1"),
panel.grid.major = element_line(linetype = "solid", size = .7, colour = "White"),
panel.grid.minor = element_line(linetype = "solid", size = .7, colour = "White")
)
# Color scales (to be manually applied to each plot)
sd <- scale_colour_manual(values = c("#88D9D6", "#990033", "#FF9966","#009999","#FFCC66","#FF9999", "#66FFFF","#3F002D"), na.value = "grey50", aesthetics = c("colour", "fill"), expand=c(0,0))
s_b <- scale_colour_manual(values = c("#990033", "grey50","#009999","#FFCC66","#FF9999", "#66FFFF","#3F002D"), na.value = "grey50", aesthetics = c("colour", "fill"), expand=c(0,0))
sgscale <- scale_colour_manual(values = c( "#88D9D6", "#3F002D","#009999"), na.value = "grey50", aesthetics = c("colour", "fill"), expand=c(0,0))
sg <- scale_fill_gradient(low = "#88D9D6", high = "#3F002D", space = "Lab", na.value = "grey50", guide = "colourbar", aesthetics = c("color", "fill") ,expand=c(0,0) )
sgnegat <- scale_fill_gradient2(low= "#88D9D6", high ="#3F002D", mid = "white",na.value ="grey50", midpoint = 0, aesthetics = c("colour", "fill"),expand=c(0,0))
# f <- ggplot() +
# geom_sf(data = geo_world,
# fill = "grey50",
# color = "transparent") +
# geom_sf(data = all_conf_geo_admin,
# aes(fill = death),
# color = "transparent") +
# labs(
# title = "Death toll is greater than zero in many provinces \nthroughout Asia, Middle East and Africa",
# subtitle = "Total fatalities per region and by type of violence \n(2016-2018)",
# caption = "Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/") +
# my_theme + sg
#
# f + coord_sf(xlim = c(-7, 140), ylim = c(-35,41), expand = TRUE)
all %>%
mutate(DATE_NUM = as.Date(EVENT_DATE, format = "%d-%B-%Y")) %>%
mutate(MONTH = month(DATE_NUM)) %>%
mutate(YR = year(DATE_NUM)) %>%
filter(YR < 2019) %>%
filter(YR > 2015) %>%
filter(EVENT_TYPE != "Headquarters or base established") %>%
group_by(EVENT_TYPE, CONTINENT) %>% summarize(Fatalities = sum(FATALITIES)) %>%
ggplot(middle_east, mapping = aes(x = fct_reorder(EVENT_TYPE, Fatalities), y = CONTINENT, fill = Fatalities), expand = FALSE) +
geom_tile() +
geom_text(aes(label=Fatalities), colour = "white", size= 3) +
coord_flip() +
labs(
x ="Type of \nconflict",
y = "Region",
title = "Most fatalities due to conflict occur as a consequence of \nwar time battles or due to remote violence",
subtitle = "Total fatalities per region and by type of violence \n(2016-2018)",
caption = "Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/") +
my_theme + sg + theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.background = element_blank(),
axis.ticks.x = element_blank(),
axis.ticks.y = element_blank())
## Warning in strptime(x, format, tz = "GMT"): unknown timezone 'zone/tz/
## 2018i.1.0/zoneinfo/America/Chicago'
The Armed Conflict Location & Event Data Project is a source has dis-aggregated data related to conflict to unprecedented levels. It presents information on violence related to combatants and violence inflicted on civilians dis aggregated by day and by type of event. The main question that is addressed by this project is who is being affected by civil conflict around the globe and what is the magnitude of it.
The graph above shows the number of fatalities by type of conflict on Asia, Africa, and the Middle East. Only data from 2016 to 2018 is used, because even though there exists data for Africa since 1990s, and data for Asia since the start of 2010, we only have data for the last three years for the Middle East. In an effort to reduce any reporting bias, I restrict to reporting information for these last three years.
Notably, we see that the two main determinants of fatalities are remote violence and battles where there is not a change of territory, and we see that it is mostly concentrated in the Middle East, consistent with the protracted conflict that Syria has experienced during the last decade. However, we see that the number of deaths as a consequence of battles is also quite high in Africa and more so in Asia, the later one surpassing the number of deaths related to other categories of war-related battles in the Middle East. It is worth noting that the amount of remote violence in the Middle East is four times the size of that in Asia, and ten times the size of that in Africa.
However, it is also relevant to underscore that where we see a highest concentration of violence outside of the battlefields, against non-combatants (civilians), is in Africa, despite the lower numbers of battle related fatalities.
This graph is aggregates information from 2016 to 2018. What are the time trends? Is violence and brutality increasing or decreasing? What is the geographic distribution within each region? How does the United States foreign aid, both economic and military, reflects this conflicts? Throughout this questions I will try to answer some of this questions.
all %>% group_by(YEAR, COUNTRY) %>%
filter(YEAR>=2015) %>%
summarise(f = sum(FATALITIES)) %>%
spread(YEAR, f) %>%
mutate(p_2018 = ((`2018`-`2017`)/`2017`)*100) %>%
mutate(t_2018 = (`2018`-`2017`)) %>%
mutate(Change = (ifelse(abs(p_2018)> 100, yes = "Change greater than 100%", no = NaN))) %>%
mutate(t_2018 = (`2018`-`2017`)) %>%
filter(p_2018!=0) %>%
filter(abs(p_2018)>20) %>%
filter(`2017` > 100) %>%
ggplot(mapping = aes(x = fct_reorder(COUNTRY, p_2018))) +
geom_bar(aes(weight= p_2018, fill = Change)) +
coord_flip() +
geom_hline(yintercept = 100, color = "red", linetype = "F1") +
geom_hline(yintercept = -50, color = "red", linetype = "1F") +
geom_hline(yintercept = 50, color = "red", linetype = "1F") +
labs(
x = "Country",
y = "Percent change in fatalities",
label = "lab",
title ="Palestine, Burkina Faso, Cameroon and Saudi Arabia \n experienced a growth in conflict of more than 100%",
subtitle = "Percent change in fatalities due to conflict between 2017 and 2018",
caption = "Methodological note: Data considers only countries \n with an initial level of over 100 deaths in 2017, following a partial definition of civil conflict \n and we only look at countries with a change greater than 20% \n\n Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/") +
s_b + my_theme %+replace%
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
legend.position = "none",
panel.background = element_blank(),
axis.ticks.x = element_blank(),
axis.ticks.y = element_blank())
Ten countries in the world have experienced a positive change of more than 20% from 2017 to 2018. Out of these, in eight countries the increase has been of 50% of more, and in four of them the increase is greater than 100%. These countries are Palestine, Burkina Faso, Cameroon, and Saudi Arabia. It is note specifying that only countries with an initial level of 100 deaths in 2017 are considered. It is also worth noting that while more than half of the countries have experienced a decrease in the conflict experienced as measured by its impact on human lives, the changes have been less stark than those moving on the different direction.
This implies that while overall trends in conflict might decrease, localized conflicts seem to be becoming more intense. The fact that fatalities have more than doubled for those four countries raise an alarm of potential prolongation and further escalation of these conflicts, and points us to the necessity of performing more research on the situation of those countries to explore potential diplomatic, regional, and multilateral actions to prevent more human loses.
per_change <- all %>% group_by(YEAR, COUNTRY, ADMIN1) %>%
filter(YEAR>=2015) %>%
summarise(f = sum(FATALITIES)) %>%
spread(YEAR, f) %>%
mutate(p_2018 = ((`2018`-`2017`)/`2017`)*100) %>%
mutate(p_2018 = if_else(p_2018 == Inf ,true = ((`2018`-`2017`)*100), false = p_2018)) %>%
mutate(p_2018 = if_else(p_2018 > 200, true = 200, false = p_2018)) #Truncating those that triplicate or more to avoid very high outliers
all_conf_geo_pchange <- left_join(geo_world, per_change, by = c('COUNTRY', 'ADMIN1'))
## Warning: Column `COUNTRY` joining factor and character vector, coercing
## into character vector
## Warning: Column `ADMIN1` joining factor and character vector, coercing into
## character vector
#sg2 <- scale_fill_gradient2(low = "green", high = "#3F002D", na.value = "grey50", guide = "colourbar", aesthetics = c("color", "fill"), breaks = "equal")
f2 <- ggplot() +
geom_sf(data = all_conf_geo_pchange,
aes(fill = p_2018),
color = "white", size = .02) +
geom_sf(data = geo_world_c,
fill = "transparent",
linetype = "solid",
colour = "white", size = .2) +
labs(
title = "Fatalities due to conflict more than doubles \n between 2017 and 2018, in several parts of the world.",
subtitle = "Percent change in deaths by conflict violence \n between 2017 to to 2018 by provinces",
caption = "Methodology note: those with an increase of over 200% are truncated to prevent \noutliers from stretching the scale \n\n Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/",
fill = "Percent change") +
my_theme + sgnegat
f2 + coord_sf(xlim = c(-9, 140), ylim = c(-35,41), expand = TRUE, datum=NA) +
theme(panel.grid.major = element_blank(),
panel.background = element_blank(),
axis.ticks = element_blank(),
axis.title = element_blank(),
axis.text = element_blank(),
legend.position = c(0.6, 0.2))
While only four countries in the world have experienced an overall increase of violence of more than 100% between 2017 and 2018, implying that fatalities more than double in these places. However, mapping the increase in violence at the first administrative level for each country, we can see that there are many regions in which the violence has not only doubled, but more than tripled between 2017 and 2018. This violence doesn’t seem to be localized, and can be found in places as diverse as several zones in the Middle East, different sub-regions in and below the Sahel stripe in Africa, Sub-Saharan Africa, and different states in South and South-East Asia including India, Bangladesh, Myanmar and Indonesia.
While there have also been places in which violence diminished between 2017 and 2018, the change has not been as pronounced in comparison.
#
# all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Government \nregains territory"] <- "Battle"
# all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-no change \nof territory"] <- "Battle"
# all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Non-state \nactor overtakes territory"] <- "Battle"
#
# sd_v1 <- scale_colour_manual(values = c("#88D9D6", "#990033", "#FF9966","#009999","#FFCC66","#FF9999", "#66FFFF","#3F002D"), aesthetics = c("colour", "fill"), expand=c(0,0))
#
# f3 <- ggplot() +
# geom_sf(data = geo_world,
# fill = "grey50",
# color = "transparent") +
# geom_sf(data = geo_world_c,
# fill = "transparent",
# linetype = "solid",
# colour = "white", size = .2)
#
#
# f3 + coord_sf(xlim = c(-10, 140), ylim = c(-35,41), expand = TRUE, datum=NA) +
# geom_point(data = all_2018,
# mapping = aes(x = LONGITUDE,
# y = LATITUDE,
# color = factor(as.factor(EVENT_TYPE))),
# size = .5,
# alpha = .3,
# position = "jitter") +
# my_theme + sd +
# guides(color = guide_legend(override.aes = list(size=3))) +
# labs(
# title = "Regional trends for all types of conflict.",
# subtitle = "Geographic distribution of events by type of violence \nin 2018",
# caption = "Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/",
# color = "Type of conflict") +
# theme(panel.background = element_blank(),
# axis.title.x = element_blank(),
# axis.title.y = element_blank(),
# legend.position = c(0.6, 0.25))
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Government \nregains territory"] <- "Battle"
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-no change \nof territory"] <- "Battle"
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Non-state \nactor overtakes territory"] <- "Battle"
all_2018_1 <- all_2018 %>% filter(EVENT_TYPE != "Remote violence") %>% filter(EVENT_TYPE != "Riots/Protests") %>% filter(EVENT_TYPE != "Violence against civilians")
sd_v2 <- scale_colour_manual(values = c("#009999", "#990033", "#FFCC66"), aesthetics = c("colour", "fill"), expand=c(0,0))
f4 <- ggplot() +
geom_sf(data = geo_world,
fill = "grey50",
color = "transparent") +
geom_sf(data = geo_world_c,
fill = "transparent",
linetype = "solid",
colour = "white", size = .2)
f4 + coord_sf(xlim = c(-10, 140), ylim = c(-35,41), expand = TRUE, datum=NA) +
geom_point(data = all_2018_1,
mapping = aes(x = LONGITUDE,
y = LATITUDE,
color = factor(as.factor(EVENT_TYPE))),
size = .5,
alpha = .3,
position = "jitter") +
my_theme + sd_v2 +
guides(color = guide_legend(override.aes = list(size=3, alpha = 1))) +
labs(
title = "Regional trends for war related conflict",
subtitle = "Geographic distribution of events by type of violence \nin 2018",
caption = "Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/",
color = "Type of conflict") +
theme(panel.background =element_blank(),
axis.title.x = element_blank(),
axis.title.y = element_blank(),
legend.position = c(0.6, 0.25))
rm(all_2018_2)
## Warning in rm(all_2018_2): object 'all_2018_2' not found
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Government \nregains territory"] <- "Battle"
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-no change \nof territory"] <- "Battle"
all_2018$EVENT_TYPE[all_2018$EVENT_TYPE == "Battle-Non-state \nactor overtakes territory"] <- "Battle"
all_2018_2 <- all_2018 %>% filter(EVENT_TYPE != "Battle") %>% filter(EVENT_TYPE != "Non-violent \ntransfer of territory") %>% filter(EVENT_TYPE != "Strategic development")
sd_v3 <- scale_colour_manual(values = c("#FF9966","#FF9999","#88D9D6","#3F002D"), aesthetics = c("colour", "fill"), expand=c(0,0))
f5 <- ggplot() +
geom_sf(data = geo_world,
fill = "grey50",
color = "transparent") +
geom_sf(data = geo_world_c,
fill = "transparent",
linetype = "solid",
colour = "white", size = .2)
f5 + coord_sf(xlim = c(-10, 140), ylim = c(-35,41), expand = TRUE, datum=NA) +
geom_point(data = all_2018_2,
mapping = aes(x = LONGITUDE,
y = LATITUDE,
color = factor(as.factor(EVENT_TYPE))),
size = .5,
alpha = .3,
position = "jitter") +
my_theme + sd_v3 +
guides(color = guide_legend(override.aes = list(size=3))) +
labs(
title = "Regional trends for non exclusively war related types of violence and unrest",
subtitle = "Geographic distribution of events by type of violence \nin 2018",
caption = "Source: ACLED dataset available at https://www.prio.org/Data/Armed-Conflict/",
color = "Type of conflict") +
theme(panel.background = element_blank(),
axis.title.x = element_blank(),
axis.title.y = element_blank(),
legend.position = c(0.6, 0.25))
rm(all_2018_1)
The events that have been reported are also found without many clear regional trends, all over the continents. As expected, riots and protests seem to be the most prevalent type of event, but we can also see a prevalence of one-sided repression throughout the African and Asian continents. It is also worth noting that there is also not a clear clustering trend within the countries, that could be expected for example if the events where to happen only around the capital of each country or around some of the bigger cities.
The two maps above are meant to be read together. One the first one, we can see a depiction of the prevalence of events categorized as an armed conflict with two or more parties fighting to secure power. The map analyzes battles, non-violent transfers of territory, and other strategic developments of the armed parties. As expected, we can see how events cluster around countries with wars typically covered by the media as Iraq, Afghanistan, and Yemen, but we can also see that there is a big prevalence of events cataloged as battles across Central Africa, South and South-East Asia. Interestingly, there are countries such as Nepal, Vietnam, South Africa, Tunisia, Algiers and Morocco, where there seem to be a lot of Strategic Developments reported, but there are not (almost) any reported battles.
As to events that are not necessarily related to wars or guerrillas (remote violence, riots and protests, and other events of violence against civilians), we see that: 1) most of the clusters of remote violence are concentrated in regions were battles are also happening (Afghanistan, Iraq, Yemen, Lebanon and Somalia for example). 2) We see a big concentration of violence against civilians that follows a similar trend to that of battles in Central Africa (a particular concentration worth noticing is Mali) and Malaysia, and we can also see that there is a big concentration in throughout the Middle East and in North-East India, bordering Nepal. 3) There is a striking prevalence of riots and protests all across India and Bangladesh, Northern Africa and Iraq and Iran.
foreign_aid %>%
filter((region_name != "World") & (fiscal_year > 1960))%>%
group_by(fiscal_year, region_name) %>%
summarise(total_raw = sum(constant_amount)) %>%
mutate(share = total_raw / sum(total_raw)) %>%
arrange(share) %>%
ggplot(aes(x=as.numeric(fiscal_year),
y = share,
fill = fct_reorder(region_name, share),
label = region_name)) +
geom_area(position = "fill")+
labs(
x = "Year",
y = "Aid share \n per \nregion",
title ="United States allocation of aid is driven largely by geo-political \n foreign policy factors, mainly military",
subtitle = "Regional distribution over time by category of United States foreign aid \n(1960-2018)",
caption = "Source: USAID dataset available at https://explorer.usaid.gov/") +
my_theme + theme(panel.grid.minor = element_line(size = 0),
panel.grid.major = element_line(size = 0),
axis.text.x = element_text(size = 8),
legend.position = "none",
panel.background = element_blank()) + sd +
geom_vline(xintercept = 1975, color = "red", linetype = "dotted") +
geom_vline(xintercept = 1979, color = "red", linetype = "dotted") +
geom_vline(xintercept = 1991, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2001, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2003, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2011, color = "red", linetype = "dotted") +
scale_x_continuous(breaks=c(1955, 1975, 1979,1991, 2001, 2003, 2011),expand=c(0,0)) +
annotate("label", y = .8, x = 1970, label = "East Asia \nand Oceania", color = "gray50", family = "Calibri", size = 3.5) +
annotate("label", y = .87, x = 1988, label = "Western \nHemisphere", color = "gray50", family = "Calibri", size = 3.5) +
annotate("label", y = .75, x = 2011, label = "South and \nCentral Asia", color = "gray50", family = "Calibri", size = 3.5) +
annotate("label", y = .7, x = 1997, label = "Europe and \nEurasia", color = "gray50", family = "Calibri", size = 3.5) +
annotate("label", y = .4, x = 2014, label = "Sub-Saharan \nAfrica", color = "gray50", family = "Calibri", size = 3.5) +
annotate("label", y = .2, x = 1990, label = "Middle East and \nNorth Africa", color = "gray50", family = "Calibri", size = 3.5) +
annotate("text", y = -.035, x = 1975 , label = "End of Vietnam \nwar ", color = "gray50", family = "Calibri", size = 3, hjust = "right") +
annotate("text", y = -.035, x = 1979, label = " Iranian \n revolution", color = "gray50", family = "Calibri", size = 3, hjust = "left") +
annotate("text", y = -.035, x = 1991, label = "End of \nCold war ", color = "gray50", family = "Calibri", size = 3, hjust = "right") +
annotate("text", y = -.035, x = 2001, label = "Start of \n Afghanistan war ", color = "gray50", family = "Calibri", size = 3, hjust = "right") +
annotate("text", y = -.035, x = 2003, label = " Start of \n Iraq war", color = "gray50", family = "Calibri", size = 3, hjust = "left") +
annotate("text", y = -.035, x = 2011, label = " End of \n Iraq war", color = "gray50", family = "Calibri", size = 3, hjust = "left")
## Warning in FUN(X[[i]], ...): NAs introduced by coercion
## Warning: Removed 6 rows containing missing values (position_stack).
When we look at the aggregated foreign aid distribution over regions and over time we observe an interesting pattern that can be tied to many historical events. During the cold war, we see an overwhelming expansion of resources into the East Asia and Oceania region, accounting for over a half of the resources for a few years, explained by the Vietnam and Korean wars and other local conflicts proxies to the soviet-american conflict. However, we see a sharp decline starting in 1980, at the same time that we see a big expansion on foreign aid invested into the Middle east and North Africa region. This corresponds chronologically to the Iranian Revolution and could be driven by efforts from the United States to have influence in the region to balance Iran’s. This focus on the MENA region persists until approximately 2005, when we see a big proportional expansion on the flow of aid to South and Central Asia, which follows the intervention in Afghanistan. Counter intuitively, we see that the share of aid to MENA declines after the war in Iraq, but this could be driven simply due to the fact that aid to Afghanistan was larger and not necessarily by a decrease in the amount of aid to the Middle East. This expansion in Central-Asia has however decreased in recent years, at the same time that we see an increase in the percentage assigned to Sub-Saharan Africa, becoming the largest recipient at this moment in time. It is worth recalling, as well, that in 2011 a big famine in Ethiopia and Somalia widely covered by the media increased the attention to the region and could be driving the increase in the share of aid to the region.
foreign_aid %>%
filter((region_name != "World") & (fiscal_year > 1960))%>%
group_by(fiscal_year, region_name, assistance_category_name) %>%
summarise(total_raw = sum(constant_amount)) %>%
mutate(share = total_raw / sum(total_raw)) %>%
arrange(share) %>%
ggplot(aes(x=as.numeric(as.character(fiscal_year)),
y = share,
fill = fct_relevel(region_name, c("East Asia and Oceania", "Western Hemisphere", "South and Central Asia","Europe and Eurasia","Sub-Saharan Africa", "Middle East and North Africa")),
label = region_name)) +
geom_area(position = "fill") +
facet_grid(assistance_category_name ~ .) +
labs(
x = "Year",
y = "Aid share \n per \nregion",
title = "United States distribution of military aid has been more volatile \nthan the distribution of economic aid",
subtitle = "Regional distribution over time by category of United States foreign aid \n(1960-2018)",
caption = "Source: USAID dataset available at https://explorer.usaid.gov/",
fill = "Region") +
my_theme + theme(panel.background = element_blank(),
axis.ticks.x = element_blank()) + sd +
geom_vline(xintercept = 1975, color = "red", linetype = "dotted") +
geom_vline(xintercept = 1979, color = "red", linetype = "dotted") +
geom_vline(xintercept = 1991, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2001, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2003, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2011, color = "red", linetype = "dotted") +
scale_x_continuous(breaks=c(1955, 1975, 1979, 1991, 2001, 2011), expand=c(0,0))
## Warning in FUN(X[[i]], ...): NAs introduced by coercion
## Warning: Removed 12 rows containing missing values (position_stack).
Breaking down the plot shown above onto two panels, we can see how these patterns are not homogeneous for military and economic aid. The trends for the economic aid are fairly stable throughout the studied years, and some trends that we can observe is somewhat of an increase of the percentage devoted to South and Central Asia, and an increase on economic aid to Europe and Eurasia after 1990. It is important to note that, however big the percentages observed for the Middle East were on the first graph presented, we can see here that it is not a place where most of the economic foreign-aid related resources go, being actually the region with the smallest percentage, throughout all the time studied.
Looking at the military pane, we see that, consistent with the observation made before, the 1960s and 1970s are marked by a big focus on aid towards East Asia and Oceania that declines after the end of the Cold-War. This is consistent with the history of the United States involvement not only on wars where it actively took part (as in Korea and Vietnam), but also on every other civic conflict that acted as a proxy for the cold-war (such as Burma, Cambodia). We also see a big investment in Europe and Eurasia which can also be related to a strategy of contingency against the Soviet influence and a stable and consistent commitment to NATO. In the mid-nineties, as the Gulf War took course we can also see the increase in the military aid towards the MENA region, and in the decade of the 2000s we can see the flow of aid towards South and Central Asia which we can infer to be related to the war on Afghanistan.
Lastly, it is important to note how there is barely any military aid towards Sub-Saharan Africa, but that nonetheless it accounts for a big part of the overall regional investments, hinting at the fact that gradually, as time has passed, most of the investments have been through channels labelled as “Economic” instead of “Military”
#Capital distance & onset
merged %>%
filter(type !=2) %>%
filter(type != "NA") %>%
ggplot(mapping = aes(x = capdist, group = onset, fill = factor(as.factor(onset), labels = c("Ongoing conflict", "Incipient conflict")))) +
geom_density(alpha = .8) +
labs(
x ="Distance to capital (kms)",
y = "Density",
subtitle = "Distribution of proximity to capital cities, according to type of conflict ",
caption = "Tollefsen, Andreas Foro; Havard Strand & Halvard Buhaug (2012) \n PRIO-GRID: A unified
spatial data structure. \n Journal of Peace Research 49(2): 363-374; \n available at https://prio.org/Data/PRIO-GRID/",
fill = "Type of conflict") +
sd + my_theme + theme(panel.background = element_rect(fill = "transparent"),
panel.grid.minor = element_line(size = 0),
panel.grid.major = element_line(size = 0),
legend.position = c(0.85, 0.4)) +
scale_x_continuous(expand = c(0,0)) +
scale_y_continuous(expand = c(0,0)) +
annotate("text", y = .0007, x = 1350, label = "Conflict Starts Close To, Or Far From,\n Capital Cities", color = "gray50", family = "Cambria", size = 6, hjust = "center")
The graph above shows the distribution of grid cells corresponding to half-degrees squares dividing up the entire globe. From these grids, I have selected only those that are located in places affected by conflict, and attempt at answering the question: Are the places where the conflict begins different than those that it spreads out to?
The graph above does not completely answer the stated question, but it does provide relevant information related to it. A kernel density graph shows us how the data is concentrated and distributed. Two main things stick out: 1) The distribution for the places where conflict started (onset) has a bi modal distribution; and 2) One of the modes of this distribution is to the left of the mode of the distribution for the other group, and the other one is to its right.
This is very interesting because it points at the fact that places where conflict first sparks are either relatively closer to the capital, and thus are potentially ignited by or with the help of urban groups and related to visible social unrest in bigger cities (or, inversely, disparities created by wealth in urban centers and their periphery ignite social unrest), or it is very far away (over 1500 km away), which might also be the reason why unrest begins (lack of acknowledgement or recognition from the government). This latter case is also in line with ideas of the literature in Conflict Studies that state that State Capacity is an important factor in preventing and controlling conflict, and the fact that the State might not have agency over remote areas can have an impact on the probability of the start of civil conflict.
merged %>% group_by(type, year) %>% filter(year> 1990) %>%
filter(type !=2) %>%
filter(type != "NA") %>%
filter(onset != 1) %>%
summarise(m1 = mean(capdist), m2 = mean(bdist1)) %>%
ggplot(mapping = aes(y = m1, x = year, group = type, color = factor(as.factor(type)))) +
geom_point() +
geom_line() +
labs(
x = "Year",
y = "Average \ndistance \nto capital (km)",
title = "Internationalized civil war has been starting closer\n to capital cities since 1996",
subtitle = "Distribution of proximity of the onset of conflict to capital cities, \n according to type of conflict.",
caption = "Source: Tollefsen, Andreas Foro; Havard Strand & Halvard Buhaug (2012) \n PRIO-GRID: A unified
spatial data structure. \n Journal of Peace Research 49(2): 363-374; \n available at https://prio.org/Data/PRIO-GRID/") +
my_theme + sd +
theme(legend.position = "none") +
geom_vline(xintercept = 1995, color = "red", linetype = "dotted") +
geom_vline(xintercept = 2003, color = "red", linetype = "dotted") +
annotate("label", y = 800, x = 2007, label = "Local", hjust = "left", color = "#88D9D6", family = "Calibri", size = 4) +
annotate("label", y = 250, x = 2007, label = "Internationalized", color = "#990033", family = "Calibri", size = 4)
To further examine the graph presented before, I take a closer look at the time-trends of the average distances to the capital cities by each type of conflict. An initial comment is that the measure for local conflicts that have not grown to become internationalized, on a first look, seems relatively steady over time although it is experiencing a slight upward trend. However, the trends in internationalized civil conflict are significantly different. We see that prior to 1995, on average the conflict happens very close to capital cities, significantly more so than non-internationalized conflicts, but then the average sharply increases between 1995 and 1996, potentially explained by better technologies to engage in more remote areas. However, after the sharp increase we can see a continuous descent, going below the threshold set by the non-internationalized civil conflict.